Single-nuclei transcriptome analysis of Huntington disease iPSC and mouse astrocytes implicates maturation and functional deficits

Summary Huntington disease (HD) is a neurodegenerative disorder caused by expanded CAG repeats in the huntingtin gene that alters cellular homeostasis, particularly in the striatum and cortex. Astrocyte signaling that establishes and maintains neuronal functions are often altered under pathological conditions. We performed single-nuclei RNA-sequencing on human HD patient-induced pluripotent stem cell (iPSC)-derived astrocytes and on striatal and cortical tissue from R6/2 HD mice to investigate high-resolution HD astrocyte cell state transitions. We observed altered maturation and glutamate signaling in HD human and mouse astrocytes. Human HD astrocytes also showed upregulated actin-mediated signaling, suggesting that some states may be cell-autonomous and human specific. In both species, astrogliogenesis transcription factors may drive HD astrocyte maturation deficits, which are supported by rescued climbing deficits in HD drosophila with NFIA knockdown. Thus, dysregulated HD astrocyte states may induce dysfunctional astrocytic properties, in part due to maturation deficits influenced by astrogliogenesis transcription factor dysregulation.


INTRODUCTION
Huntington disease (HD) is a devastating, autosomal-dominant neurodegenerative disease characterized by movement abnormalities, psychiatric disturbances, and cognitive impairment with no disease-modifying treatment. The genetic cause of HD is a CAG repeat expansion in the first exon of the huntingtin (HTT) gene, which encodes an expansion of a polyglutamine (polyQ) repeat tract in the huntingtin (HTT) protein. 1 When the HTT CAG repeat is expanded to 40 CAGs or above, HD is fully penetrant. 2,3 HTT is ubiquitously expressed in all cell types in the body and a broad range of cellular processes, predominantly in the brain, are impacted by chronic mutant HTT (mHTT) expression. 4,5 The mutation alters HTT structure and function, causing toxic gain of function and loss of normal HTT functions. 4 While the most overt HD neuropathology involves striatal neuronal loss and cortical atrophy, modeling of human-induced pluripotent stem cells (iPSCs) has been leveraged to identify HD cell-autonomous defects in non-neuronal cell types, such as endothelial cells that comprise the blood-brain barrier, 6 oligodendrocytes, 7 and astrocytes, [7][8][9][10] indicating that chronic expression of mHTT may induce intrinsic deficits within each cell type.
Astrocytes are the major glial cell type in the CNS, playing a critical role in regulating brain homeostasis. Astrocytes provide neurotrophic support through glutamate uptake and recycling, and contribute to synaptogenesis, neuronal maturation, and neuronal maintenance. Inter-astrocytic communication of calcium signaling is a function of astrocytes thought to help control blood flow for coupling to neuronal energy demand. 11, 12 In addition to regulating a microenvironment that facilitates neuronal signaling, astrocytes maintain direct interactions with the brain endothelium to establish and maintain BBB and neuronal functions through soluble and physical communications. 13 There has been growing literature characterizing the development and regulation of unique cellular states that define functional states of astrocytes. Moreover, impairments in these astrocyte states may generate toxic functions that contribute to neurological disorders; however, the extent and mechanisms driving these impairments are just emerging for HD. HD astrocyte dysfunction appears to contribute to neuronal dyshomeostasis, [14][15][16] including impaired glutamate signaling and K+ buffering in astrocytes from HD mice. 15,[17][18][19] Glutamate-related gene expression changes and dysfunction have recently been demonstrated in human patient postmortem tissue 19 and human patient iPSC-derived cells, 9 respectively. HD iPSC-derived astrocytes carrying juvenile-onset range CAG repeat lengths (77 CAG and 109 CAG) provided less support for functional neuronal maturation and contributed to glutamate-induced toxicity in iPSC-derived striatal neurons. 9 Other HD iPSC astrocyte studies revealed an increased autophagic response 8 and increased evoked inflammatory responses. 20 Additionally, purification of striatal astrocytes revealed common changes to RNA expression from R6/2 and Q175 HD mouse models, including core genes in calcium-dependent processes, G protein-coupled receptors, and glutamate receptor signaling 19 or in pathways regulating cytoskeletal formation and cell branching. 10 R6/2 uniquely showed changes to cholesterol regulation. 10 However, studies have yet to examine the molecular heterogeneity at a singlecell level to define distinct astrocyte states from striatum and cortex in HD mouse models and compare these to patient iPSC-derived astrocytes, as well as define potential mechanisms controlling their dysregulation.
With the development of single-cell and single-nuclei RNA-sequencing (sc-or snRNA-seq), astrocyte cell state transitions can now be compared at high resolution across multiple species, brain regions, and ages. A recent snRNA-seq study on HD astrocytes from postmortem patient cortical tissue identified heterogeneous cell states in HD human astrocytes. 21 Increased expression of metallothionein and heat shock genes and loss of astrocyte-specific gene expression was demonstrated. To further investigate HD human astrocyte cell states, human patient iPSC-derived astrocytes allow the ability to ascertain earlier, cell-autonomous events at the single-cell level that may lead to later changes in postmortem human tissue. Furthermore, iPSC-derived astrocytes allow for functional assessment, linking transcriptional states to astrocyte dysfunction. Findings from this model system can be compared to in vivo mouse and other HD model systems to evaluate common and novel HD phenotypes.
The goals of the present study are to further characterize astrocyte cell states in unaffected and HD contexts, identify drivers of transcriptional dysregulation, and to compare these states across human iPSC and mouse model systems. We investigated which astrocyte states exist in HD and if their transcriptional signatures can help predict altered astrocyte functions. We used unbiased snRNA-seq clustering followed by pathway enrichment and systematically compared signaling states between human and mouse astrocytes. Using this approach, we found altered astrocyte cell states related to glutamate signaling and maturation common to both HD iPSC-derived astrocytes (iAstros) and R6/2 astrocytes from the striatum and cortex compared to unaffected control astrocytes. We also identified disease cell states that predict activated extracellular matrix (ECM) and dysregulated actin cytoskeletal dynamics to be unique to human HD astrocytes, informing cell-autonomous dysregulation related to endogenous CAG repeat expansion in astrocytes. These transcriptional signatures were then used to identify regulatory genes that may be driving these alterations and could represent therapeutic targets. Overall, HD transcriptional changes reveal potential astrocyte maturation deficits and implicate that astrogliogenesis transcription factors, including ATF3 and NFIA, play a regulatory role contributing to HD pathogenesis. Together, these data provide a scientific resource of astrocyte cell states and transitions in both iPSC and mouse models of HD that may be useful for broadening understanding of the diverse astrocytes in the brain and guide therapeutic intervention in HD. The results suggest that loss of identity and maturation impairments are driven by altered expression of specific astrogliogenesis transcription factors.

HD iAstros exhibit altered morphology and astrocyte marker expression
To understand how CAG repeat expansion may influence astrocyte function in HD, we evaluated heterogenous astrocyte cell states through quantifiable gene expression analysis at the single-cell level. We first investigated whether altered cell state changes are reflected through cell-intrinsic mechanisms in astrocytes differentiated from human HD versus control iPSCs that express mHTT in an endogenous context. To do so, HD (46Q, 53Q) and control (18Q and 33Q) 6,23,24 iPSCs were differentiated into astrocytes (iAstros) (Table S1). Briefly, iPSCs were subject to prolonged neural induction to generate gliogenic neural stem cells followed by exposure to astrogliogenic factors for 60 days to induce morphologically and functionally mature astrocytes (Figures 1A, S1, and Table S2). HD iAstros had fewer astrocytic processes and had (B) At day 60, GLAST+/DAPI-fluorescence-activated cell sorting (FACS) was performed for astrocyte enrichment and astrocytes are subsequently matured for 7 days. At day 67, GLAST + astrocytes underwent single-nuclei isolation prior to droplet-enabled single-nuclei RNA-sequencing (snRNA-seq). (C) Representative FACS histogram of HD and control iAstros GLAST expression level demonstrates HD astrocytes have decreased expression of GLAST at day 60. (D) FACS quantification of GLAST+/DAPI-populations in day 60 iPSC-derived astrocytes (one-way ANOVA: 18Qv46Q ****p < 0. iScience Article significantly larger cell volumes ( Figure S2), indicating a potential deficit in astrocyte maturation. 25 We observed a significant decrease in the percentage of HD iAstros compared to control astrocytes that express GLAST, a functional glutamate transporter that clears the excitatory neurotransmitter from the extracellular space at synapses and is expressed by mature human astrocytes ( Figures 1C and 1D). To directly compare a more pure population of HD to control mature astrocytes, GLAST-positive iAstros were isolated by fluorescence-activated cell sorting (FACS) and used for all subsequent studies ( Figures 1B-1D and S3). GLAST-sorted HD and control cells express canonical astrocyte markers, like GFAP, and exhibit a classical star-like morphology ( Figure 1F).
To further investigate morphological differences, we performed Sholl analysis on GLAST-sorted HD and control iAstros. HD GLAST-positive iAstros had less total number of astrocytic processes ( Figure 1G) compared to GLAST-positive control iAstros. Additionally, HD GLAST-positive iAstros had shorter average length of processes ( Figure 1H). Together, decreased length and number of astrocytic processes indicate that GLAST-positive HD astrocytes may have a decreased maturation profile compared to GLAST-positive control astrocytes.
Astrocyte transcriptional states in HD patient-derived iAstros reveals immature cell states regulated by aberrant astrogliogenesis transcription Next, to investigate astrocyte transcriptional signatures in GLAST-sorted HD and control iAstros, we used snRNA-seq ( Figure 1B). Among the 36,128 nuclei sequenced, nine distinct cell clusters were identified across HD and control iAstros representing unique astrocyte cell states across Uniform Manifold Approximation and Projection (UMAP) plots (Figures 2A and S4). HD or control clusters were identified by the number of iAstros per genotype by cluster ( Figure 2B). Cluster 0 highly expressed mature astrocyte markers ( Figure 2C) and represented a large proportion of control iAstros. Clusters 4 and 6 highly expressed other astrocyte markers, like S100b and CD44; these clusters were largely composed of control iAstros, with cluster 4 being primarily composed of a single control iAstro cell line (CTR33Q) ( Figure S4). Clusters with a larger proportion of HD iAstros (e.g. clusters 1, 2, 3, 5, and 8) expressed low levels of mature astrocyte markers and/or high levels of immature human astrocyte markers (MKI67 and TOP2A), as transcriptionally defined from fetal human astrocytes. 25 In addition to expression of immature astrocyte markers, cluster 8 had high expression of progenitor markers, like PAX6, and had a larger population of HD cells. Cluster 7 also highly expressed immature astrocyte markers but showed similar proportions by genotype. Based on gene expression profiling of astrocyte developmental markers, control and HD iAstros reflect most cell states including various developmental states; however, HD iAstros show clear shifts in composition and full depletion of one of the control states, suggesting altered astrocyte maturation.
To molecularly assess the unique transcriptional signatures across iAstro clusters, the top markers per cluster were plotted and classified into gene categories ( Figure 2D). iAstro clusters 0 and 6 had a large proportion of control iAstros and had the highest expression of neural support-related markers, like astrocyte-secreted synaptogenesis regulator SPARCL1 and brain-derived neurotrophic factor receptor NTRK2. Immune-related genes, including CD74 and many HLAs, were also highly expressed in iAstro cluster 6. Cluster 4 was composed of almost entirely control iAstros and represented a unique cell state with the most highly expressed genes related to long non-coding RNAs, like NEAT1, and reduced expression of ECM and progenitor markers. Clusters 1 and 2 contained a majority of HD iAstros and had high expression of adhesion-related molecules, LGALS3 and PALLD, with cluster 2 also highly expressing calcium-binding molecules, ANXA2 and SLC3A1. Immature astrocyte markers and cell cycle-related genes were highly expressed among a cluster with a large proportion of HD iAstros (cluster 3). Cluster 5 contained the largest proportion of HD iAstros and showed increased expression of several ECM-related genes, including collagens and fibronectin, along with calcium-binding and adhesion-related genes, which were also highly expressed in clusters 1 and 2 that contained a large proportion of HD iAstros. Cluster 8 was the smallest cluster (493 nuclei; 1.4% of total nuclei sequenced) and had a gene expression profile consistent with astroglial progenitors. Cluster 7 had equivalent proportions of HD and control iAstros and had a relatively high expression of DNA repair and transcription-related genes, like FANCA and EZH2. Overall, control states were enriched for astrocyte functions involved in neural support and neuroinflammation, while HD states were enriched for adhesion, ECM, and immature markers. These data highlight a loss of normal astrocyte functional support, potential developmental impairments, and ECM and adhesion-related alterations in HD iAstros. Finally, there is complete loss of a cell state in HD iAstros that showed high expression of NEAT1, a lncRNA regulator of transcription that has previously been implicated in HD 26,27 and is involved in altered neural cellular development and neural cell damage. 28 Figure 2. Dysregulated HD patient-derived astrocytes implicate immature cell states regulated by aberrant astrogliogenesis transcription (A) GLAST + iAstro snRNA-seq UMAP by genotype (CTR n = 2, HD n = 2) and by cluster. (B) Genotype composition of cells across astrocyte clusters and by genotype. (C) Astrocyte marker genes across each iAstro cluster shows a decrease in maturation gene expression in several subclusters with larger proportion of HD cells (teal defined by >60% of the cluster were CTR cells, pink defined by >60% of cluster were HD cells, gray genotype represents 40-60% CTR or HD cells). (D) Top cluster markers for each cluster for cell state classification. (E) Top most significant pathways across all iAstro clusters used to classify cell state signatures. Psuedobulk genotype pathway enrichment using DEGs from all HD iAstros compared to control iAstros (Data S1) is included for comparison. (F) Top 5 most activated and inhibited significant pathways in HD iAstros generated using genotype DEGs (Data S1). See also Figures S4 and S7. To further explore the functional relevance of the iAstro clusters to biological pathways, ingenuity pathway analysis (IPA) pathway enrichment was performed on the genes differentially expressed between clusters. The top 2 most significant pathways per cluster were compared to assess the transcriptional signatures unique to each iAstro cluster and to differentially expressed genes (DEGs) by genotype (pseudobulk) ( Figure 2E and Data S1). Hepatic fibrosis/hepatic stellate cell activation was a pathway among the top two significantly enriched pathways in all iAstro clusters except 3 and 6. Cluster 0, which had the largest proportion of control iAstros, exhibited significant inhibition of kinetochore metaphase signaling (z = À3.16, p < 3e-9), composed of cyclin, centromere, and histone-related molecules (CCNB1, CDK1, CENPE, CENPK, H2AX, and H2AZ1). Kinetochore metaphase signaling was instead activated in the immature cluster 3 (z = 4.64, p < 2e-43), that had a larger number of HD iAstros. The significant enrichment of hepatic fibrosis/hepatic stellate cell activation (z = 0, p < 1e-5) largely due to the downregulation of several collagens (COL1A1, COL4A1, and COL4A2) was present in cluster 4 that contained nearly all control iAstros. Another cluster that had a large proportion of control iAstros (cluster 6) had significant activation of neuroinflammatory signaling molecules (z = 3.74, p < 6e-14), like HLA transcripts and SLC1A2. The cluster with the second largest proportion of HD iAstros (cluster 1) had significant enrichment of antigen presentation pathway (z = 0, p < 3e-7) due to the expression of RNA transcripts that encode various HLA proteins (HLA-A, HLA-C, HLA-DPA1, HLA-DRB5, and HLA-E). The significant upregulation of several collagen transcripts (COL11A1, COL18A1, COL1A2, COL27A1, COL3A1, COL4A1, COL4A2, and COL6A2) caused a cluster with a large HD population (cluster 2) to be enriched in GP6 signaling molecules (z = 0, p < 7e-8).
Together, HD iAstro gene expression changes by cluster and by genotype highlight an inhibited glutamate receptor signaling state and unique astrocyte cell states, such as actin cytoskeletal signaling activation and immature astrocytes, that may contribute to loss of neural support and altered morphology or motility.
Defining HD striatal and cortical mouse astrocyte cell states Human HD iAstros show shifts in cell state compositions, including complete loss of a state seen in control iAstros (cluster 4), that represents mature astrocytes with decreased expression of ECM-related genes and increased expression of NEAT1. We noticed a shift in the composition of HD iAstros that indicated an abundance of immature and progenitor-like cells before and after GLAST-positive sorting; these differences may represent deficits in astrogliogenesis which may lead to the functional state changes seen above. We next compared if similar state changes occur in an in vivo animal model of HD, and whether speciesspecific states exist. To assess the transcriptional cell states of striatal and cortical mouse astrocytes within a well-established system that recapitulates many transcriptome changes in human HD tissue, 30 we carried out snRNA-seq analysis of striatal and cortical brain regions from the rapidly progressing R6/2 HD model, iScience Article generated through the transgenic overexpression of the first exon of human HTT, and non-transgenic (NT) mice at ages 8 weeks (symptomatic) and 12 weeks (highly symptomatic) 31,32 ( Figures 3A and 4A).
At the 12-week time point, we identified 13 and 16 distinct clusters from 26,509 striatal and 25,237 cortical nuclei sequenced, respectively. Expression of the Slc1a2 astrocyte marker gene that encodes GLT1, a glutamate transporter that clears the excitatory neurotransmitter from the extracellular space at synapses, was used to identify the astrocyte cluster for each dataset ( Figures 3B and 4B). The cluster with the highest Slc1a2 expression was subset to investigate R6/2 astrocyte-specific dysregulation via sub-clustering and further transcriptomic analyses. Due to a small population of cells with relatively low levels of astrocyte markers and high expression of vascular markers Pdgfrb and Flt1, 33 additional sub-setting was necessary before sub-clustering the cortical astrocyte cluster ( Figure S5). Upon sub-clustering all 12-week R6/2 astrocyte clusters, we identified six striatal astrocyte clusters and five cortical astrocyte clusters visualized using UMAP ( Figures 3C and 4B). Interestingly, the total number of clusters identified was less than the human iAstros which may represent species differences, considering the known complexity of human glia versus mouse glia. 34 Diverse cellular states were represented across each mouse dataset and to identify genotype composition by cluster, the total number of cells per genotype was quantified by cluster. Striatal clusters 2 and 4 were significantly enriched in R6/2 compared to NT, while cortical cluster 0 was significantly enriched in R6/2, indicating these clusters are relatively novel to the R6/2 condition ( Figures 3D and 4C). R6/2 striatal astrocytes had a larger number of differentially expressed genes by genotype compared to cortical astrocytes ( Figures S6 and S7).
Expression of mature astrocyte marker genes and transcription factors regulating astrogliogenesis were visualized across astrocyte clusters ( Figures 3E and 4D). R6/2-enriched striatal astrocyte clusters 2 and 4 as well as R6/2-enriched cortical cluster 0 had a lower expression of several classic astrocyte identity genes, such as Slc1a2 and glutamate-ammonia ligase (Glul), compared to NT-enriched striatal cluster 1. Similarly, cortical astrocyte clusters 1 and 3 had a lower expression of these astrocyte identity genes compared to cluster 0. Decreased expression of astrocyte markers in R6/2-enriched clusters indicates that dysregulated cell states exist in R6/2 astrocytes and may reflect differences in astrocyte development or a loss of cell identity genes. Altogether, there was a shift in the R6/2 astrocytes toward more immature cell states.
Inhibited synaptogenesis-related signaling states in R6/2 striatal astrocytes We next molecularly assessed the cell states identified in striatal and cortical R6/2 and NT mouse astrocytes. First, the top highly expressed gene markers per cluster were used to classify clusters into functionally relevant categories ( Figures 3F and 4E). Several of the highest expressing genes in significantly NTenriched striatal cluster 1 and cortical cluster 2 included astrocyte-enriched glutamate transporter (Slc1a2) and glutamate receptors (Gria1, Gria2, and Grin2c), like iAstro clusters 0 and 6 that had larger proportion of control cells. The striatal clusters with the largest proportion of NT astrocytes (clusters 1 and 3) also had high expression of GABA transporter Slc6a11 and GABA receptor Gabbr2. Cortical clusters 1 and 2 had a high expression of an actin-binding gene Clmn and a phagocytic receptor Mertk. Cortical cluster 4, which had a large proportion of NT astrocytes, exhibited high expression of neural regulatory genes (Anks1b and Nrg3), axonal/synaptic-related genes (Robo2, Dlg2, and Lrtm4), and Kcnip4, which encodes a potassium channel-interacting protein.
Among the striatal clusters with significantly more R6/2 cells compared to NT, striatal clusters 2 and 4 had the highest expression of developmental-related genes, RNA polymerase (Polr2a), and a nuclear receptor involved in neurogenesis (Nr6a1). R6/2-enriched striatal cluster 2 also had the highest expression of two ECM-related genes: glycoprotein-related Sgcd and cell adhesion molecule Nrxn3. R6/2-enriched cortical cluster 0 highly expressed Ogt, a glycosyltransferase regulator. Similar to iAstro clusters 1 and 2 that had large proportion of HD iAstros, striatal cluster 5 had the highest expression of sodium, potassium, or calcium ion-binding genes, like Adgrv1, Nkain2, and Cacnb2, except for the potassium channel, Kcnd2, which was most highly expressed in striatal clusters 1 and 2. Cortical cluster 3 exhibited high expression of calcium-binding genes, Adgrv1 and Cacnb2, and was not enriched by genotype.
Cortical cluster 2, which was largely composed of NT astrocytes, and non-genotype-enriched cluster 1 were significantly enriched in glutamate receptor signaling due to expression of Grin2c and Slc1a2, with predicted activation in cluster 1 (z = 1, p < 5e-8) but no predicted activation for cluster 2 (z = 0, p < 0.002). Cortical cluster 1 also exhibited a predicted activation of synaptogenesis signaling (z = 2.83, p < 0.0006). Glutamate receptor signaling was the most significantly enriched pathway for cortical cluster 3, with a predicted inhibition (z = À1.34, p < 4e-11), like HD iAstros. Additionally, Rho GDP-dissociation inhibitor (RhoGDI) signaling was predicted to be most activated and synaptogenesis signaling was the most inhibited predicted pathway for non-genotype-enriched cortical cluster 3.
Overall, the data suggest a loss of glutamate signaling and synaptogenesis-related pathways across R6/2 astrocytes, and activation of ECM and immature-like astrocyte states, similar to human HD patient iAstros. These data highlight similarities with the human HD iAstros that suggest maturation impairments, with a particular focus on glutamate receptor and synaptogenesis-related dysregulation.

HD mouse astrocyte temporal and brain region comparisons
We next examined transcriptional changes that occur in R6/2 astrocytes with progressive disease. We assessed the activation scores of the top ten most significant pathways enriched in 8-week mouse astrocyte signaling states and 12-week mouse astrocyte signaling states compared to age-matched NT mouse striatal astrocytes ( Figure 3H and Data S2). As expected, there was overlap of many relevant pathways between the 8-and 12-week striatal R6/2 astrocytes. The 12-week astrocytes exhibited increased predicted activation of iCOS-iCOSL signaling in T helper cells (12-week: z = 1, p < 0.008; 8-week: z = 0, p < 0.032), PI3K signaling in B lymphocytes (12-week: z = 0.447, p < 0.003; 8-week: z = 0, p < 0.048), and neuroinflammation signaling (12-week: z = 0.447, p < 5e-5; 8-week: z = 0, p < 0.001). Chuk, which encodes an inhibitor of the transcription factor NFkB complex, was an overlapping upregulated gene across the three inflammatory pathways. The most significantly inhibited pathway in 12-week striatal astrocytes was cAMP-mediated signaling (z = À1.67, p < 3e-5). Interestingly, cardiac hypertrophy signaling was the most significantly inhibited pathway predicted in the 8-week striatal R6/2 astrocytes    iScience Article (z = À0.447, p < 0.008) due to expression of genes such as Camk2g, Chuk, and Fgf14, but was predicted to be activated in 12-week striatal R6/2 astrocytes (z = 0.302, p < 6e-4). In general, the most significantly enriched pathways in 8-week striatal astrocytes had no predicted activation or inhibition. By 12 weeks, R6/2 astrocytes had more severe predicted activation or inhibition of these overlapping pathways that correlates with increased severity in motor phenotypes, 35 suggesting increased astrocyte dysfunction with disease progression. The overlap in top pathways between 8 and 12 weeks with progressive dysregulation of similar genes, indicates that at these two symptomatic timepoints, R6/2 astrocytes are similarly dysregulated. Taken together, R6/2 astrocytes exhibited inhibition of neuronal homeostatic signaling pathways, suggesting how mHTT-expressing striatal astrocytes could contribute to neuronal dysregulation in R6/2 mice.
Temporal changes occurring in R6/2 cortical astrocytes were also evaluated ( Figure 4G and Data S2). The most significantly enriched pathways in 12-week cortical astrocytes were predicted to be inhibited or had no activation scores. Among those were glutamate receptor signaling (z = 0, p < 8e-5), synaptic longterm potentiation (z = À2.236, p < 2e-4), and neuroinflammation signaling pathway (z = 0, p < 2e-4), with glutamate receptor Grin2c being a common downregulated gene across the three pathways. The most significantly enriched pathway in 8-week cortical astrocytes was glutamate receptor signaling (z = 0, p < 0.004), which was also significantly enriched in 12-week cortical astrocytes due to downregulation of related genes (Glul, Gria2, Grin2c, and Slc1a2). The 8-week cortical astrocyte pathways had no predicted activation or inhibition (z = 0), like 8-week striatal astrocytes. These findings further demonstrate dysregulated signaling in R6/2 astrocytes related to neuronal homeostasis, particularly glutamate receptor signaling (which overlaps with results from HD iAstros), increases in severity with disease pro-gressionÀalthough diversity in cell states was less severe in R6/2 cortical astrocytes compared to R6/2 striatal astrocytes.
To determine the impact of mHTT on R6/2 astrocyte cell states by brain region, we directly compared striatal and cortical astrocyte pathway enrichment performed on genotype DEGs from both timepoints ( Figures 5A, S5, and Data S2). R6/2 12-week striatal and cortical astrocytes showed significant inhibition of neural regulatory-related signaling, including synaptic long-term potentiation, synaptogenesis, and CREB signaling in neurons. The most significant pathways enriched in R6/2 8-week striatal and cortical astrocytes had low or no predicted activation scores, but the individual pathways were more unique to each brain region. When evaluating each region, striatal 8-week top pathways included neuroinflammation and GABA receptor signaling, while cortical 8-week top pathways were glutamate receptor signaling, ceramide degradation, and sphingosine-1-phosphate (S1P) metabolism; this highlights the unique cellular processes in metabolic regulation across these two brain regions during this stage in HD pathogenesis that may drive changes at later stages. There were many overlapping significant pathways relating to synaptogenesis signaling in 12-week striatal and cortical timepoints but had increased predicted inhibition in cortical astrocytes. These shared and unique pathways across astrocytes in age and brain region highlight the unique astrocyte cell states that may contribute to R6/2 pathogenesis across these two highly affected brain regions.
As the overlapping pathways across time points and brain regions suggested, there was significant astrocyte genotype DEG overlap ( Figure 5B and Data S2). The set of 56 overlapping 12-week striatal and cortical DEGs were genes all dysregulated in the same direction and may represent common signatures in astrocytes independent of region ( Figure 5C). Individual examination of each of these genes was performed to uncover commonalities contributing to transcriptional dysregulation in R6/2 mouse astrocytes at this severe disease stage. Gria2 and Grin2c were present in a majority of the overlapping dysregulated signaling  iScience Article pathways related to neuronal support, glutamate receptor signaling, and ion binding; this highlights how aberrant R6/2 astrocytes affect the regulation of neuronal homeostasis. There were also several genes related to both neural support and extracellular matrix/adhesion-related signaling, such as Sparcl1, Ttyh1, and Tspan7. Together, this dysregulated astrocyte signaling at 12 weeks may highlight two major dysfunctions occurring across R6/2 astrocytes and how it affects the cells they are meant to support, therefore contributing to pathogenesis.
To further assess the common dysregulation across R6/2 astrocytes, we examined the nine genotype DEGs that were common across astrocytes from both brain regions and time points. These genes all had very similar expression levels that often became more severe with age, with striatal 12-week astrocytes typically having the most extreme expression of these nine common DEGs ( Figure 5D). Exceptions included several splicing and transcription-related genes (Gm3764, Tra2a, and Ddx5) that had the highest fold change in DEGs from cortical 12-week R6/2 astrocytes compared to cortical NT astrocytes. A non-coding RNA (Gm4876) and a lipid regulator (Abca1) were DEGs unique to only glial cell types (astrocytes, oligodendrocytes, and/or oligodendrocyte progenitor cells) in the larger R6/2 snRNA-seq dataset. Among these overlapping astrocyte DEGs, the downregulation of the glycogen phosphorylase activator, Phkg1, was unique to only astrocytes compared to other R6/2 cell types and highlights the potential dysregulation of glycogen breakdown in R6/2 astrocytes. These common DEGs may represent core processes, metabolic and transcriptional, that become dysregulated at an early stage of HD pathogenesis and give rise to other changes seen at 8 and 12 weeks in the R6/2 mice. The changes in these metabolic and transcriptional processes may lead to the altered signaling described above related to synaptogenesis, glutamate signaling, calcium signaling, and inflammation.

Inhibition of glutamate signaling in HD iAstros and R6/2 astrocytes
A significantly inhibited glutamate receptor signaling cell state was identified in R6/2 and HD iAstro clusters. The presence of these states in both the mouse and iPSC model suggests that this is a core feature of HD astrocytes and likely arises intrinsically from the expression of mHTT. Genes that encode glutamate transporters (SLC1A2 and SLC1A3) and glutamate receptors (GRIA1, GRIA2, and GRIN2A) were significantly downregulated across both models of HD astrocytes (Figures 6A and 6B; Data S1 and S2). To confirm the intrinsic loss of glutamate signaling in HD astrocytes, we investigated protein levels of a representative glutamate receptor 1 (GluR1), encoded by GRIA1. Western blot analysis for GluR1 was performed on HD and control iAstros and showed significantly decreased GluR1 in HD iAstros compared to control iAstros ( Figure 6C). Due to the common downregulation of glutamate-related signaling seen in both mouse and human HD astrocytes in this study and in previously published results, 50-52 we next assessed if our HD iAstros showed a functional impairment. HD iAstros demonstrated decreased functional uptake of glutamate when exogenous glutamate was added in vitro ( Figure 6D). Since the downregulation of glutamate transporters and receptors was observed in both models, and we demonstrated decreased functional glutamate uptake in HD iAstros, we suspect that this is a cell-autonomously downregulated signaling pathway in HD astrocytes. The sorted iAstros are not exposed to signaling from neurons, and recapitulation of this alteration provides evidence for cell-autonomous effects. Inhibition of the uptake of glutamate can influence HD pathogenesis by contributing to glutamate neurotoxicity.

Activation of ECM and actin in HD iAstros
ECM and actin cytoskeletal signaling were processes uniquely activated in a subcluster of HD iAstros which were not identified in R6/2 astrocytes. While there were several actin, ECM, and adhesion-related DEGs (e.g. CADM1, CLMN, and MICAL1) present in overlapping 12-week striatal and cortical R6/2 astrocytes ( Figure 5C), this did not trigger significant enrichment of specific actin or ECM signaling pathways, unlike HD iAstros. We investigated the genes contributing to ECM and actin cytoskeletal signaling activation in the HD and control iAstro cell lines to determine the genes controlling these dysregulated pathways (Figure 7A and Data S1). Genes related to extracellular signaling (ITGB1 and ITGA3) and those regulating actin  Figure 7B and Data S1). Given the potential disruption of the actin cytoskeleton, we investigated whether filamentous actin (Factin) protein expression was increased in our HD iAstros as a surrogate readout. Phalloidin in-cell westerns were performed on HD and control iAstros and showed that HD iAstros had significantly increased expression of F-actin compared to control iAstros ( Figure 7C). Additionally, HD iAstros exhibited increased levels of the F-actin crosslinking protein, actinin (ACTN), compared to control iAstros ( Figure 7D), suggesting increased actin cytoskeletal scaffolding as a gain of cell state in HD iAstros. Taken together, ECM and actin cytoskeletal signaling are cell-autonomous and potentially early-activated signaling pathways unique to patient-derived HD astrocytes that may contribute to morphological changes or increased adhesion in HD astrocytes.

Aberrant astrogliogenesis transcriptional regulators in HD astrocytes
To investigate the potential regulatory mechanisms controlling human HD astrocyte dysregulation, we used genotype DEGs in HD iAstros (Data S1) to predict transcription factor regulation enrichment analysis by chromatin immunoprecipitation-X enrichment analysis (ChEA) 36,37 ( Figure 8A). The top predicted transcriptional regulator was an important regulator of astrogliogenesis, ATF3, followed by transcriptional repressor ZNF217 ( Figure 8A). To validate this finding, we looked at protein levels of ATF3 by Western blot. Compared to control iAstros, HD iAstros had significantly decreased ATF3, suggesting that aberrant HD astrogliogenesis iScience Article may be driven by known transcriptional regulators of this process ( Figure 8B). Furthermore, SOX9, a transcriptional regulator of glial development and early astrogliogenesis, which was downregulated in HD iAstros, also had significantly decreased protein levels by Western blot ( Figure 8C). The neural stem cell factor SOX2 was also enriched. YAP1 of the HIPPO signaling pathway that has an essential role in regeneration and self-renewal was another significantly enriched factor in HD iAstros that has previously been implicated in HD and shown to interact with Htt. 38 Overall, HD astrocyte transcriptionally dysregulated cell states and predicted alterations in transcription factors may implicate an astrogliogenesis deficit that induces astrocyte dysfunctions in glutamate signaling and activation of ECM and actin cytoskeletal signaling to potentially contribute to HD pathogenesis via altered glutamate uptake and cellular motility, respectively.
To identify possible transcriptional regulators in R6/2 astrocytes, we conducted transcription factor enrichment analysis 37 using the 56 common 12-week DEGs ( Figure 5C and Data S2). The ChEA database showed glial developmental transcription factor, SOX9, as the most significant transcription factor enriched, suggesting a role for this complex in altered gene expression in R6/2 mouse astrocytes ( Figure 8D). TCF4, a transcription factor involved in nervous system development, and SUZ12, a member of the polycomb repressive complex 2, were also enriched, suggesting roles for these complexes in repressing  iScience Article astrogliogenesis genes in HD ( Figure 8D), that have also been implicated in HD neurons, 39 endothelial genes in HD BMECs, 6,40 and HD oligodendroglia. 40 Another predicted transcriptional regulator in R6/2 astrocytes was the transcriptional repressor ZNF217, also present in HD iAstro analysis (Figures 8A and 8D). SOX9 protein expression was evaluated in the striatum of 10-week-old R6/2 mice; however, there was no significant change in expression levels or number of SOX9-expressing astrocytes (GFAP-positive or S100b-positive cells) (Figures 8E and 8F). For further investigation into transcription factor perturbation in R6/2 mouse astrocytes, Enrichr transcription factor database was queried ( Figure S6B). Hes family bHLH transcription factor 1 (HES1) overexpression was implicated as the most significant perturbation, further implicating transcriptional repression of nervous system development in R6/2. Alterations in ASCL1 and MECP2, regulators of neural development, were also predicted to contribute to common dysregulated gene sets in R6/2 striatal and cortical astrocytes. NFIA, a transcription factor that controls astrogliogenesis, was identified as being deficient, indicating a potential mechanism for altered astrocyte development in R6/2. These findings highlight potential mechanisms in astrogliogenesis and neural developmental dysregulation contributing to downstream pathway dysregulation in mHTT-expressing astrocytes-a similar finding to our human iAstro model of HD.

Modulation of astrogliogenesis regulator NFIA/Nfl in HD drosophila glia is protective
Lastly, evaluated if perturbation of an astrogliogenesis transcriptional regulator can rescue HD astrocyteinduced deficits in vivo. The Drosophila model system was leveraged given the ease of genetic manipulation, short life cycle, and simple functional assays used to assess complex glial biology 43,44 as well as previous data showing a deficit in glia in HD model flies. 41,42 An HD fragment Drosophila model where mHTT (HTT 231NT128Q ) is expressed in a pan-glia (repo) driver was crossed with lines that have a knockdown of the fly ortholog to human NFIA, NfI. We used a general driver of mHTT in glia versus an astrocyte-specific driver, such as Alarm-Gal4, because of the robust phenotypes previously reported using repo-Gal4 42 and that a large proportion of cells targeted by the repo driver are astrocytes. 45 The astrogliogenesis transcriptional regulator, NfI, was evaluated due to its high orthology prediction (DIOPT) score 46 with human NFIA and high activity in the larval ventral nerve chord, suggesting it may have an orthologous function in flies. Using a climbing assay, 41 mHTT flies exhibited a significant deficit in climbing compared to control flies ( Figure 8G). The mHTT-induced climbing deficit was improved in lines with NfI knockdown in all cells (NfI KD) or in only glia (NfI glia KD) by 7 days and rescued to control levels by 10 days ( Figure 8G). Therefore, one of the astrogliogenesis transcriptional regulators predicted from dysregulated HD mouse and human astrocyte transcriptomic analysis demonstrates that perturbation of astrogliogenesis can improve HD functional deficits in vivo.

DISCUSSION
Unbiased single-cell approaches define cell states across the landscape of thousands of individual cells, providing a unique window into the cell states that arise during disease. 47 Cell types whose function span a wide range of biological processes, such as astrocytes, are ideal candidates for the investigation of cell state transitions in disease. The goal of our studies was to identify and compare astrocyte transcriptional cell states in specific systems and time points, their transitions in unaffected control and HD systems, and to elucidate the regulatory signaling involved. We investigated transcriptome signatures from human iPSC-derived astrocytes and astrocytes from a transgenic mouse model to elucidate common and speciesspecific cell states, as well as aberrant cell states that exist in both human and mouse HD astrocytes. Both models demonstrated potential loss of cell states involved in neural support, glutamate signaling, ionbinding, and neural development. Both also suggest astrocyte developmental alterations that are predicted to be regulated by astrogliogenesis transcription factors. HD human astrocytes exhibited a unique activation of ECM and cytoskeletal signaling that was not observed in R6/2 astrocytes. While most of the transcriptional cell states we identified in both human and mouse data were represented in HD and control samples, with shifts in percent composition in each condition, the HD iAstro data revealed complete loss of a unique control cell state that showed high expression of NEAT1, a gene associated with the stability of nuclear paraspeckles and thought to function through interactions with RNA binding and other proteins and RNAs, thus affecting transcription. 28,48 NEAT1 has also been implicated in striatal neurons from patients with HD and mouse models, with its overexpression protecting from mHTT-induced cytotoxicity. 27 The R6/2 astrocytic dysregulation that overlapped between brain regions and with the human data showed an increase in severity between 8 and 12 weeks, suggesting that astrocyte dysregulation is progressive as described previously for bulk RNA-seq analysis of purified astrocytes 19 ; however, there were unique differences found in 8-week cortical astrocytes involving SP1 signaling and lipid metabolism, which may ll OPEN ACCESS iScience 26, 105732, January 20, 2023 iScience Article represent unique age-dependent cortical changes or early changes that give rise to the dysregulation of related molecular processes at 12 weeks. Overall, the dysregulated astrocytic states identified here are anticipated to induce dysfunctional brain homeostasis and contribute to HD pathogenesis.
Single-nuclei technology has allowed us to identify diverse astrocyte states within a heterogeneous astrocyte population through unsupervised computational clustering that may not have been identified using traditional bulk transcriptomic analysis. 49 Homeostatic astrocyte states identified from control iAstros and NT mouse astrocytes, compared to HD iAstros or R6/2 mouse astrocytes, included activation of signaling pathways previously suggested to play roles in astrocyte functions. Synaptogenesis signaling and CREB signaling were significantly activated in NT striatal and cortical astrocyte states along with calcium-binding activation in the cortex. In control human iAstros, several synaptogenesis-related genes were among the top upregulated cluster markers (SPARCL1) in control iAstro state cluster 0. Similarly, glutamate receptor signaling was significantly activated across an NT striatal mouse astrocyte state relative to R6/2 striatal astrocyte clusters. Overall, snRNA-seq analyses of control iAstros and non-transgenic mouse astrocytes were enriched with terms implicating synaptogenesis, glutamate signaling, and other neural supportive transcriptional cell states that are typical of homeostatic astrocytes.

HD human iAstro and R6/2 astrocyte phenotype comparisons
The commonly dysregulated astrocyte state across HD human iAstros and R6/2 mouse astrocytes was the inhibition of glutamate signaling. This phenotype is consistent with previous studies suggesting that glutamate handling is a critical feature of excitotoxicity in HD. [50][51][52] Astrocytes terminate the action of neurotransmitters, like glutamate and gamma-aminobutyric acid (GABA), through uptake at the tripartite synapse. In the case of glutamate, astrocytes help clear this excitatory neurotransmitter from the synaptic cleft through transporters (GLT1 and GLAST) and with the help of transmembrane glutamate receptors (GluR1 and GluR2) that form ligand-gated ion channels. [53][54][55] Within the cytosol of astrocytes, glutamate is then degraded into the neuro-inactive state, glutamine, via glutamine synthetase (GLUL) in an ATP-dependent manner; each of these were downregulated in our present study. These changes were validated by decreased protein levels of the glutamate transporter GLAST, in human HD iAstros upon FACS isolation, and AMPA glutamate receptor GluR1 in GLAST-positive HD iAstros compared to control. These alterations lead to functional impairment though decreased glutamate receptor signaling, which correlated with decreased glutamate uptake in HD iAstros. This dysregulation is present in R6/2 and other models of HD and may involve potassium ion channel Kir4.1 in R6/2 mice. 9,16,18,19,[56][57][58] The evidence of dysfunctional glutamate regulation in our adult-onset HD iAstro lines is consistent with prior studies demonstrating a similar phenotype in juvenile-onset HD iPSC-derived astrocytes, 9 and with loss of GLT1 expression in mice with HTT-160Q selectively expressed only in astrocytes. 56 Our findings further support a cell-autonomous component of aberrant glutamate signaling inhibition in HD astrocytes with functional consequences in human cells.
Axonal guidance was an inhibited pathway predicted in one of the iAstro cell states composed largely of HD cells and overlapped with striatal R6/2 astrocyte cell states. This neural developmental pathway is necessary to guide neurons to their correct targets for proper synaptogenesis; therefore, inhibition of this critical process would alter neuronal circuitry homeostasis. Moreover, both iAstro cluster 3 and R6/2-enriched striatal cluster 4 appeared to be in a more immature, proliferative cell state. Aberrant activation of cell cycle-related signaling has been demonstrated in HD mice, 35,59 HD iPSC-derived neural models, 23,24,60,61 and HD patient tissue. 62,63 Osipovitch et al also demonstrated a decrease in GFAP expression in patient with HD embryonic stem cell-derived astrocytes that reflects impaired astrocytic differentiation. 7 Developmental alterations in HD iAstros and decreased mature astrocyte markers in a striatal R6/2 astrocyte subpopulation may be a common cell state associated with mHTT-induced decreased astrogliogenesis. Transcription factor analysis and protein validation in HD iAstros identified downregulation of early astrogliogenesis factors: SOX9 and ATF3. 64,65 These genes may be driving maturation impairments in HD astrocytes, and subsequently causing alterations in astrocyte functions as these immature cell states also showed decreased expression of functional genes. HES1 overexpression was one of the most significantly enriched transcription factors in R6/2 12-week astrocytes along with the dysregulation of several HES genes in HD iAstros; this is consistent with epigenetic dysregulation of a HES family transcription factor (HES4) associated with striatal degeneration in HD postmortem cortex. 66  iScience Article phenotypes may also be due to a loss of astrocyte identity rather than a developmental alteration, as suggested in a previous study 67 ; however, the human HD iAstros data suggest a developmental phenotype given the decreased propensity for HD iPSCs to differentiate into GLAST-positive, morphologically mature astrocytes, similar to altered in vivo astroglial differentiation of human glial progenitor cells from HD embryonic stem cells. 7 Both hypotheses could be occurring as HD progresses and mature cells lose their identity while progenitor cells give rise to developmentally impaired astrocytes. Nonetheless, alterations in astrocyte functions are likely, in part, due to the lack of mature astrocyte characteristics. Rescuing these deficits may require the proper induction of transcriptional regulators that are involved in the development and maturation of astrocytes and the expression of mature functional genes. Alternatively, persistent activation of developmental regulators may impede proper differentiation and maturation of HD astrocytes, which would require appropriate downregulation of these genes.
NFIA is a master astrogliogenesis transcriptional regulator that acts as a co-regulator with SOX9 in early astrogliogenesis 68 and ATF3 in later astrogliogenesis. 65 We assessed perturbation of NFIA in HD model animals. When the fly ortholog to human NFIA, NfI, was knocked down in either all cells or in glia of a fly model of HD, the functional capacity for the mHTT repo flies to climb was significantly improved. An interpretation of this is that if a downregulated gene is knocked down and the phenotype is rescued, that gene may be downregulated as a compensatory mechanism. 42 Another interpretation may be due to the differences in timepoints assessed across the three HD models utilized in this study. The autophagic balance of these transcription factors, like many others in HD, may be altered differently depending on disease stage due to mHTT-induced alterations in clearance and accumulation of such proteins. [69][70][71] To this account, SOX9 was significantly decreased in HD iAstros and predicted to be a transcriptional regulator, but was not significantly decreased in R6/2 astrocytes. Investigations into additional timepoints across HD models are necessary for elucidation of this hypothesis. Overall, the astrogliogenesis transcriptional factors, like NFIA, are dysregulated across multiple HD models and can be targeted for improved functional outcomes in a model of HD.
Astrocyte reactivity has been documented in HD mouse [72][73][74][75] and human tissue 76,77 ; however, the extent of cell-autonomous astrogliosis has not been extensively characterized. Surprisingly, control iAstros appeared to exhibit a more activated neuroinflammatory state compared to HD; however, this was a relatively small cell state in iAstro cluster 6, only encompassing 5% of the total nuclei sequenced. This suggests that control human iAstros may have a greater ability to mount normal inflammatory responses, including those that are a hallmark of in vitro cultured astrocytes. 78 Neuroinflammatory signaling was enriched in R6/2 striatal and cortical genotype-specific differentially expressed genes, with minor activation in striatal 12-week astrocytes; however, looking closely into the cell states, cortical cluster 4 and iAstro cluster 6 had a predicted activation of neuroinflammation signaling and had populations that shifted toward unaffected control or NT cells. Consistent with this hypothesis, R6/2-enriched striatal cluster 4, which had the lowest expression of mature astrocyte markers, had a predicted inhibition of neuroinflammatory signaling, demonstrating a potential loss of cell function in HD. These data suggest that both activated and inhibited neuroinflammatory states may exist in HD and could be playing different roles in the disease. Some of these neuroinflammatory states may only arise due to extrinsic signaling from other CNS cell types, as suggested by the lack of neuroinflammatory states in the HD iAstros. Since immune responses of astrocytes can induce detrimental and/or beneficial effects on neighboring cell types, it is unknown what consequences would result from loss of this cell state. 79 While a lack of inflammatory response in HD astrocytes contrasts with evidence in literature showing evidence of reactive astrocytes in HD mouse models [72][73][74] and human postmortem tissue, 57,72,[75][76][77] more recent studies support a lack of traditional or A1-like reactive astrogliosis states in HD. 15,16,19,21 Overall, there were several dysregulated pathways common to HD human iAstros and R6/2 astrocytes. Differences between systems may be due to multiple reasons, including expression of an expanded repeat HTT exon 1 transgene in R6/2, while HD iAstros contain the full-length endogenous mHTT. Recent studies identified similarities and differences across astrocytes from truncated (R6/2) and full-length (zQ175) mHTT mouse and human stem cell-derived models 10 or postmortem human striatal tissue. 19 Some similarities across mouse and human models in these studies and our study highlighted ion homeostasis, metabolism, neurotransmitter signaling (including glutamate receptor signaling), and astrocyte morphology. 10,19 Finally, species differences may also contribute as human astrocytes have greater development of astrocyte networks both in numbers and complexity, exhibiting significantly more multi-branched processes ll OPEN ACCESS iScience 26, 105732, January 20, 2023 19 iScience Article than rodent astrocytes. 12,34 The increased ability of adult human astrocytes to respond to extracellular glutamate compared to mouse astrocytes is an example that suggests adult human astrocytes may have evolved an improved capability to respond to synaptic activity. 25,34,80 Unique HD human iAstro phenotypes Signaling states unique to human HD iAstros included activation of actin cytoskeletal and integrin signaling. Additionally, HD iAstros had a decrease in matrix metalloproteinases (MMP15 & 17) that aid in ECM degradation. Altered actin cytoskeletal dynamics was validated by increased Phalloidin (F-actin) staining and the actin crosslinking protein, ACTN, in HD iAstros at the protein level. This gainof-function cell state unique to human HD iAstros compared to R6/2 astrocytes suggests a cellular state transition that is most likely a cell-autonomous effect, not influenced by the presence of other cell types that provide developmental cues for astrocytes. Altered actin cytoskeletal dynamics in HD have been implicated due to interaction with HTT, including the major actin monomer sequestering protein, profilin. [81][82][83] The functional relevance of actin and integrin activation to disease pathogenesis has yet to be determined but may relate to altered morphology, adhesion, intracellular trafficking, and transcriptional regulation.

Unique R6/2 astrocyte dysregulated cell states
Cyclic adenosine monophosphate (cAMP)-mediated signaling was the most inhibited pathway in striatal and cortical R6/2 astrocytes at 12 weeks. This signaling pathway plays a vital role in cellular signaling, including the regulation of glucose and lipid metabolism. Phkg1 was a downregulated gene common across both timepoints and both brain regions, and was unique to R6/2. Since astrocytes are the primary storage site for glycogen in the brain, 84 a decrease in Phkg1 would reduce glycogen phosphorylase's ability to break down glycogen in astrocytes needed for the high energy demand of neurons. Decreased glucose levels have been shown in HD astrocytes purified from mouse striatum. 85 Furthermore, decreased striatal metabolism in HD patients has been attributed to decreased glucose uptake through GLUT3. 86,87 In addition to disrupting neuronal activity, reduced glucose levels may impact downstream epigenetic and transcriptomic signatures.
We also described cellular states unique to the two brain regions assessed. Striatal R6/2 cell state differences highlighted changes in maturation and ion transport/binding astrocytic processes, while cortical R6/2 astrocytes mostly presented glycosylated cell state transition with an inhibited synaptogenesis signature. Striatal R6/2 astrocytes and HD iAstros clusters had a loss of astrocyte identity genes. This molecular phenotype was less overt in cortical R6/2 astrocytes with fewer significant differentially expressed genes across genotype for 12-week and 8-week cortical astrocytes compared to striatal, which yielded less diversity in significant pathway dysregulation. The differenes in molecular phenotype by brain region may correspond with more overt neurodegeneration in the striatum, compared to the cortex.
In summary, our findings indicate that aberrant cell states exist in individual HD astrocytes that include inhibition of glutamate signaling and potential developmental alterations that may be regulated by mutant HTT's interaction with astrogliogenesis transcription factors, ATF3, NFIA, and SOX9. In addition, loss of astrocyte maturation may contribute to inhibition of astrocytic functions, such as glutamate signaling, in human HD and R6/2 astrocytes. Knockdown of NFIA in HD flies rescued motor deficits and may suggest that restored protein levels of this key astrogliogenesis transcription factor would be neuroprotective, possibly by allowing glial cells to reach full maturity if the protein is functioning in the HD context. The identification of common and unique HD cell states across a heterogeneous population of human and mouse astrocytes provides directions for further mechanistic evaluation and greater information relating to potential therapeutic interventions targeting these dysregulated astrocytic states in HD as well as predicted outcomes across model systems.

Limitations of the study
In this study, iPSC-derived astrocytes were grown in vitro to mimic conditions of human astrocyte growth; however, this methodology results in early-stage astrocytes in 2D culture, therefore may lack signals and physical interaction from other neurovascular unit cell types that they would be exposed to in vivo. As discussed above, we interpret the HD iAstro results as potential cell-autonomous features of bona fide HD human astrocytes, and future studies are aimed at further investigating this hypothesis. Secondly, it is possible that differences in genetic background between HD and control patients may have confounded iScience Article the analysis across the non-isogenic HD iAstro lines analyzed. Nonetheless, we noted consistent dysregulated features across HD iAstros that were lacking in unaffected control iAstros, such as glutamate signaling downregulation and actin-related signaling upregulation, with some of these features recapitulated in HD mouse astrocytes. Additionally, male and female control iPSC lines were used in this study, while all male HD patient-derived iPSC lines were used for much of the analysis due to efficiency of differentiations and availability of lines. The data in this study were based on readouts from one to two timepoints and may not have captured subtle developmental and maturation-related features of HD astrocyte dysregulation. Lastly, the HD models assessed represent various disease stages and disease induction. For example, the R6/2 mouse is a rapidly progressing model of HD induced by the overexpression of the expanded CAG repeat-containing first exon of the mHTT gene, causing transcriptional deficits and severe motor dysfunction by 10-12 weeks of age and is representative of later stages of manifest disease. On the other hand, HD iAstros are generated from a pluripotent state in patient cells with endogenous full-length mHTT and are in a more developmentally immature state. Taken together, our results demonstrate astrocyte maturation deficits in multiple HD models that may influence dysregulated HD astrocyte cell states to induce dysfunctional astrocytic properties.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS
iPSC lines and cell culture HD and unaffected control iPSCs were generated via non-integrating episomal reprogramming of patientderived fibroblasts (see Table S1). 6,[22][23][24] The iPSC lines were characterized by morphological analysis and immunocytochemistry (see Table S3) for presence of pluripotency markers and lack of differentiation markers. All iPSC lines exhibited normal karyotypes by aCGH array performed by Cell Line Genetics (Madison, WI). The five iPSC lines were maintained on Matrigel hESC-qualified matrix (Corning) with mTeSR1 (STEM CELL Technologies). Once 60-80% confluent, iPSCs were passaged using a 5-10 minute incubation with Versene (Gibco) at 37 C for expansion purposes.

Astrocyte differentiation
Five iPSC lines were differentiated into neural progenitor cells (NPCs) according to manufacturer's instructions (Gibco #A1647801, MAN0008031). NPCs were cultured to 90-95% confluency, then seeded at 1e5/cm 2 on hESC-qualified matrigel (Corning) for four passages before induction of astrocyte differentiation. All NPC lines exhibited normal karyotypes by aCGH array performed by Cell Line Genetics (Madison, WI) and expressed NPC/NSC markers as assessed by immunocytochemistry. To begin astrocyte differentiation, passage four NPCs were seeded at 7.5e4 cells/cm2 in Neural Expansion Medium (Gibco #A1647801) and the following day (day 0), half the medium was replaced with astrocyte differentiation medium (ADM) 1. Half media changes were performed every other day for the remainder of the differentiation. All subsequent passages were performed by washing cells with HBSS without Mg Ca (Gibco) three times, then incubating in StemPro Accutase (Gibco) at 37 C for 10-15 minutes, counting using TC20 Automated Cell Counter (Bio-Rad) with trypan blue exclusion, and seeding on hESC-qualified matrigel (Corning) at 5e4 cells/cm 2 unless otherwise noted. On day 7, the cells were replated in ADM1. On day 15, the cells were replated in 50% ADM1, 50% ADM2. Subsequent half media changes were performed using ADM2. On day 30, day 34, and day 45, the cells were seeded in 100% ADM3 at 4e4 cells/cm 2 . 88 On day 60, the cells were subject to astrocyte enrichment via FACS. Quality control immunocytochemistry was performed at each passage time point for PSC, NPC, neural, and astrocytic markers. See Tables S2 and S3.

R6/2 mouse model
All experimental procedures were in accordance with the Guide for the Care and Use of Laboratory Animals of the NIH and animal protocols were approved by Institutional Animal Care and Use Committees at the University of California Irvine (UCI), an AAALAC accredited institution. For this study, five-week-old R6/2 31 and non-transgenic (NT) male mice were purchased from Jackson Laboratories and housed in groups of up to five animals/cage under a 12-hour light/dark cycle with ad libitum access to chow and water. Mice were aged to 8, 10, or 12 weeks then euthanized by pentobarbital overdose and perfused with 0.01 M PBS. Striatum and cerebral cortex were dissected out of each hemisphere and flash-frozen for snRNA-seq (8 and 12 weeks) or immunohistochemistry (10 weeks).

Fly husbandry
Fly lines were generated as described 89 or ordered from the Bloomington Drosophila Stock Center (BDSC) or Vienna Drosophila Stock Center (VDSC). A 231 amino acid N-terminal fragment of mHTT with 128 glutamines was expressed using a pan-glial (repo) driver. This line was generated in the Botas lab at Baylor ll OPEN ACCESS iScience Article room temperature. Cells were incubated with Phalloidin-iFluor790 (Abcam, ab176763) at 1:1000 for 1 hour at room temperature, followed by washing three times with 0.1% Tween 20 (Sigma-Aldrich) in PBS. CellTag700 (LI-COR, 926-41090) at 1:500 was added as a total protein stain (for normalization purposes) and incubated for 1 hour at room temperature. Cells were then washed three times with 0.1% Tween 20 (Sigma-Aldrich) in PBS, then solution was completely removed from wells. Wells were immediately imaged using the LI-COR Odyssey CLx and analyzed using LI-COR's Empiria Studio Software.

Western blots
Day 67 GLAST-positive astrocytes were broken in 150uL modified RIPA buffer, sonicated three times for 10 seconds at 40%, then spun down for 20 minutes at 4 C, 1000 rpm. Bradford (Thermo Scientific BioMate 3S UV-visible spectrophotometer) protein quantification was performed with cuvettes. Protein lysates were aliquoted at 5 mg for GluR1, ACTN, ATF3, and SOX9 (see Table S3). Then equivalent amounts of loading buffer were added. Gels were run at 150 V using Boltä 4-12% Bis-Tris Plus Gels, 15-well gels with MOPS, transferred onto Nitrocellulose Membrane 0.45 mm at 10 V for 1 hour at room temperature. The membrane was rinsed in water and then let dry for 15 minutes. Membranes were rehydrated in water and then followed the Revert 700 Total Protein Stain protocol. Membranes were blocked in Intercept (TBS) Blocking Buffer for 1 hour at room temperature. Membranes were incubated in primary antibody (GluR1, ACTN, ATF3, and SOX9) and rotated at 4 overnight. Membranes were then washed three times with TBST, then incubated in secondary antibody: GluR1 into IRDyeâ 800CW Goat anti-Rabbit IgG, ACTN into IRDyeâ 800CW Goat anti-Mouse IgG, ATF3 into IRDyeâ 800CW Goat anti-Mouse IgG, and SOX9 in IRDyeâ 700CW Goat anti-Rabbit IgG for 1 hour at room temperature. Membranes were washed three times with TBST while gently shaking for 1 hour, then washed two times with TBS at room temperature. Membranes were imaged using the LI-COR Odyssey CLx and analyzed using LI-COR's Empiria Studio Software.

Single-nuclei RNA-sequencing Mouse
Single nuclei from three mice per genotype per timepoint were isolated from half hemisphere striatum or cortex in Nuclei EZ Lysis buffer (Sigma-Aldrich, NUC101-1KT) and incubated for 5 minutes. Samples were passed through a 70 mm filter and incubated in additional lysis buffer for 5 minutes and centrifuged at 500 rcf for 5 minutes at 4 C before two washes in Nuclei Wash and Resuspension buffer (1x PBS, 1% BSA, 0.2 U/ mL RNase inhibitor). Nuclei were FACS sorted using DAPI to further isolate single nuclei and remove additional cellular debris. These nuclei were run on the 10x Genomicsâ Chromium Single cell 3' gene expression v3 platform. Libraries were quality controlled and sequenced on the Illumina NovaSeq 6000.
iAstros Day 67 GLAST-positive cells were thawed and seeded into hESC-qualified matrigel coated 6-well plates at 4e4 cells/cm 2 in ADM3 (see astrocyte differentiation). After 48 hours, day 69 GLAST-positive cells were washed with HBSS without Mg Ca (Gibco) three times then lifted using a 10-15 minutes incubation of StemPro Accutase (Gibco) at 37 C. Single-cell suspensions were strained using a 70 mm strainer, washed using cold 0.08% BSA (Gibco) in PBS, and counted using TC20 Automated Cell Counter (BioRad) with trypan blue exclusion per manufacturer's instructions in the 10x Genomicsâ Demonstrated Protocol: Single Cell Suspensions from Cultured Cell Lines for Single Cell RNA Sequencing (Document CG00054). To obtain a single-nuclei suspension, cells were lysed using Nuclei EZ Lysis Buffer (Sigma Aldrich, NUC101-1KT) for 5 minutes on ice. Lysed cells were centrifuged 500 rcf for 5 minutes at 4 C, then washed by twice resuspending in 1% BSA (Gibco), 0.2 U/mL of Human Placenta RNase Inhibitor (NEB), in PBS and centrifuged again. After two washes, the nuclei were strained using a 40 mm cell strainer, counted using the TC20 Automated Cell counter (Bio-Rad), resuspended at 1000 nuclei/mL in 0.08% BSA (Gibco) in PBS, then proceeded immediately with the 10x Genomicsâ Chromium Single cell 3' gene expression v3 platform. Libraries were quality controlled and sequenced on the Illumina NovaSeq 6000.

Data processing Mouse
A total of 109,053 cells with 6.1 billion reads were sequenced for the 24 samples (on average 4,544 cells per sample with $56K reads per cell). Alignment was done using the CellRanger pipeline v3.1.0 (10x Genomics) to a custom pre-mRNA transcriptome built from refdata-cellranger-mm10-1.2.0 transcriptome using ll OPEN ACCESS iScience Article Fly climbing assay Virgin female flies harboring the mHTT transgene or control flies (VDSC 13974) were crossed to males harboring experimental alleles. Experimental crosses were raised and maintained at 25 C. Only female progeny were assessed using the climbing assay. Anesthetized using carbon dioxide, female progeny were sorted into vials used for experimentation without food in groups of 9 to 10 flies with 10 replicates (biological replicates). The percent of flies that climbed to or past 5 cm in 10 seconds at room temperature were quantified. Tapping was used to elicit negative geotaxis to force flies to the bottom of the vials with 3 rapid succession taps, three times, over 2 seconds. Flies were allowed to climb for 10 seconds, and a picture was taken against a white background with a mark parallel to the lab bench at 5 cm from the bottom of the vial. After a 1-minute recovery, the 10 second climbing trial was repeated for a total of five times (technical replicates). Flies were climbed at days 7 and 10 post-eclosion.

QUANTIFICATION AND STATISTICAL ANALYSIS
All features highlighted in the paper and reported as statistically significant have p-values < 0.05 or adjusted p-values (FDR) < 0.1, unless otherwise stated.

ll
OPEN ACCESS